Leaders today increasingly turn to big data and advanced analytics in hopes of solving their most pressing problems, whether it’s a drop-off of repeat customers, a shift in consumption patterns or an attempt to reach new markets.
What works with locomotives and oil rigs, however, can be far less effective when it comes to influencing people’s behaviors. With social systems and the behaviors generated by large groups of individuals — who does what and under what conditions — it is far harder to identify solutions to problems. This points to the shortcoming of using data analytics alone for solving problems that arise from individual behavior. Here are five important considerations that everyone who works with big data needs to understand:
— DATA CAN DETERMINE THE “WHAT” OF A PROBLEM: Data analysis is helpful in determining patterns of behavior, both positive and negative — for example, the success of an organization or enterprise in motivating people to engage in certain activities.
— DATA RARELY REVEAL THE “WHY”: In the aggregate, individual behaviors show up in the data, revealing patterns among certain demographics and groups. Data may prompt people to make assumptions, though they are only guesses about the rationale of others’ behaviors, not a reliable basis for determining the best solution to address a problem.
— THE “WHY” NEEDS A QUALITATIVE APPROACH: Whether the social group involves current customers, potential customers, vendors or any other population, the only way to discover the “why” is to engage with them in qualitative research such as interviews, focus groups and observation.
— CONSIDER TEMPORAL AND OTHER FACTORS: They also influence behavior, making solutions more difficult to find and less likely to remain effective over time.
— RIGOROUS TESTING CAN FIND THE RIGHT SOLUTION: With big data analysis and smaller-scale qualitative research combined, organizations can gain deeper insights into both problems and their causes, which can then help inform solutions that are likely to produce a desired result. The best way to know the effectiveness of a solution is to conduct randomized testing using two similar groups: one that is offered the solution and one that is not.
